simplify with directive

[wanghan02]

1. when used as x3::with(x), where x is an lvalue, we inject an lvalue reference of x to the context.
2. when used as x3::with(x), where x is a const lvalue, we inject a const lvalue reference of x to the context.
3. when used as x3::with(x), where x is an rvalue, we move the rvalue to the member of with_directive and inject an lvalue reference of that member to the context.
4. No copy is applied; std::ref can still be used but not needed.
5. remove unused with_context

[djowel]

Looks good! Can you provide some test files?

[wanghan02]

@djowel Test cases have been added to test/x3/with.cpp
rvalue injection: The 1st commit actually inject a const ref to the member variable because parse(...) is a const method. This behaves the same as before (but without copying), but does not really make sense. I think the best way to use with directive is:

• Inject non-const lvalue: able to modify during parsing, the original lvalue is modified.
• Inject rvalue: able to modify during parsing.
• Inject const lvalue: not able to modify during parsing.

So I brought back the with_value_holder and change the T const specialization to T& specialization, and makes rvalue injection modifiable. Test cases have been changed to adapt to the new requirement.

[djowel]

wonderful!

[Kojoley]

Could you explain why is it mutable?

1. The tests you had added do not cover it.
2. Allowing to mutate the value creates thread-safety risks.
3. It does not play well with constexpr, especially with Clang 6 and below.

[wanghan02]

The idea was rvalue injection should be mutable in const parser as well. Given parsers are immutable in the parse functions, rvalue injection would be immutable without mutable keyword.

But I also agree with you on the issues it brings. I don't have a solution at the moment.